Preserving Task-Relevant Information Under Linear Concept Removal

• URL: http://arxiv.org/abs/2506.10703v1
• Date: Thu, 12 Jun 2025 13:53:29 GMT
• Title: Preserving Task-Relevant Information Under Linear Concept Removal
• Abstract summary: SPLICE eliminates sensitive concepts from representations while preserving their covariance with a target label.<n>It is the unique solution that removes linear concept predictability and maintains target covariance with minimal embedding distortion.<n> Empirically, SPLICE outperforms baselines on benchmarks such as Bias in Bios and Winobias.
• Score: 19.953631890863498
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Modern neural networks often encode unwanted concepts alongside task-relevant information, leading to fairness and interpretability concerns. Existing post-hoc approaches can remove undesired concepts but often degrade useful signals. We introduce SPLICE-Simultaneous Projection for LInear concept removal and Covariance prEservation-which eliminates sensitive concepts from representations while exactly preserving their covariance with a target label. SPLICE achieves this via an oblique projection that "splices out" the unwanted direction yet protects important label correlations. Theoretically, it is the unique solution that removes linear concept predictability and maintains target covariance with minimal embedding distortion. Empirically, SPLICE outperforms baselines on benchmarks such as Bias in Bios and Winobias, removing protected attributes while minimally damaging main-task information.

Related papers

• Soft Symbol Grounding for Prototypical Concepts [1.9584254864124988]
  textbfSoft-PNet reframes concept grounding as a Metropolis walk over a cache of feasible symbolic solutions.<n>It trains against one KL objective between the prototype-weighted cache and the network's concept predictions.<n>On textttMNIST-EvenOdd, Visual Sudoku, and textttKand-Logic under scarce supervision, Soft-PNet matches loss-engineered networks at the concept and label levels and recovers concepts that soft-grounding baselines miss.
  arXiv  Detail & Related papers  (2026-09-10T22:10:33Z)
• EraseSAE: Surgical Concept Erasure in Text-to-Video Diffusion Models via Sparse Autoencoders [103.46709005379591]
  Concept erasure offers a principled remedy by removing unwanted semantics from pretrained models.<n>Existing approaches typically operate at a coarse quality, misaligned with the finegrained, distributed nature of concept representations.<n>We propose EraseSAE, a novel framework that leverages sparse autoencoders to achieve surgical concept erasure.
  arXiv  Detail & Related papers  (2026-09-03T10:23:37Z)
• To Erase, or Not to Erase: Robust Training-Free Concept Erasure with Preservation aware Adaptive Ranked Subspace Expansion [10.950549540413528]
  Concept erasure techniques edit text-to-image diffusion models to erase undesired targets such as NSFW content or copyrighted styles.<n>Current CETs face a trade-off between erasure and utility: stronger edits erase the target more reliably but degrade utility on non-target concepts.<n>We present Preservation-aware Adaptive Ranked Subspace Expansion (PARSE), a training-free framework for robust concept erasure in latent diffusion models.
  arXiv  Detail & Related papers  (2026-07-26T06:35:31Z)
• Erasing Without Collateral Damage: Precise Concept Removal in Diffusion Models [10.269961654607108]
  Training-free concept erasure is attractive mechanism for controlling text-to-image diffusion models.<n>But precise erasure often comes at the cost of damaging semantically related non-target concepts.<n>We introduce CARE, a closed-form concept erasure operator that replaces the raw target direction with a kept-subspace-aware direction.
  arXiv  Detail & Related papers  (2026-07-06T16:21:09Z)
• Beyond Heuristic Prompting: A Concept-Guided Bayesian Framework for Zero-Shot Image Recognition [81.2779530670268]
  Vision-Language Models (VLMs) have significantly advanced zero-shot image recognition.<n>This paper enhances prompt by incorporating class-specific concepts.<n>Our method consistently outperforms state-of-the-art approaches.
  arXiv  Detail & Related papers  (2026-03-09T03:11:11Z)
• Localized Concept Erasure in Text-to-Image Diffusion Models via High-Level Representation Misdirection [7.863370847022858]
  We propose High-Level Representation Misdirection (HiRM), which misdirects high-level semantic representations in the text encoder toward designated vectors.<n>Our strategy enables precise concept removal with minimal impact on unrelated concepts.<n>HiRM also preserves generative utility at low training cost, transfers to state-of-the-art architectures such as Flux without additional training, and shows synergistic effects with denoiser-based concept erasing methods.
  arXiv  Detail & Related papers  (2026-02-23T09:18:27Z)
• MeGU: Machine-Guided Unlearning with Target Feature Disentanglement [73.49657372882082]
  We propose a novel framework that guides unlearning through concept-aware re-alignment.<n>MeGU enables controlled and selective forgetting, effectively mitigating both under-unlearning and over-unlearning.
  arXiv  Detail & Related papers  (2026-02-19T05:20:31Z)
• Differential Vector Erasure: Unified Training-Free Concept Erasure for Flow Matching Models [49.10620605347065]
  We propose Differential Vector Erasure (DVE), a training-free concept erasure method specifically designed for flow matching models.<n>Our key insight is that semantic concepts are implicitly encoded in the directional structure of the velocity field governing the generative flow.<n>During inference, DVE selectively removes concept-specific components by projecting the velocity field onto the differential direction, enabling precise concept suppression without affecting irrelevant semantics.
  arXiv  Detail & Related papers  (2026-02-01T08:05:45Z)
• Geometric-Disentangelment Unlearning [106.99160454669902]
  gradient ascent on forget samples often harms retained knowledge.<n>We propose the Geometric-disment Unlearning (GU) that decomposes any candidate forget gradient update into tangential and normal components to retain space and executes only the normal component.<n>Our method is plug-and-play and can be attached to existing gradient-based unlearning procedures to mitigate side effects.
  arXiv  Detail & Related papers  (2025-11-21T09:58:25Z)
• Does Representation Intervention Really Identify Desired Concepts and Elicit Alignment? [73.80382983108997]
  Representation intervention aims to locate and modify the representations that encode the underlying concepts in Large Language Models.<n>If the interventions are faithful, the intervened LLMs should erase the harmful concepts and be robust to both in-distribution adversarial prompts and the out-of-distribution jailbreaks.<n>We propose Concept Concentration (COCA), which simplifies the decision boundary between harmful and benign representations.
  arXiv  Detail & Related papers  (2025-05-24T12:23:52Z)
• CURE: Concept Unlearning via Orthogonal Representation Editing in Diffusion Models [6.738409533239947]
  CURE is a training-free concept unlearning framework that operates directly in the weight space of pre-trained diffusion models.<n>The Spectral Eraser identifies and isolates features unique to the undesired concept while preserving safe attributes.<n>CURE achieves a more efficient and thorough removal for targeted artistic styles, objects, identities, or explicit content.
  arXiv  Detail & Related papers  (2025-05-19T03:53:06Z)
• SPEED: Scalable, Precise, and Efficient Concept Erasure for Diffusion Models [56.83154571623655]
  We introduce SPEED, an efficient concept erasure approach that directly edits model parameters.<n>Speedy searches for a null space, a model editing space where parameter updates do not affect non-target concepts.<n>We successfully erase 100 concepts within only 5 seconds.
  arXiv  Detail & Related papers  (2025-03-10T14:40:01Z)
• UPCORE: Utility-Preserving Coreset Selection for Balanced Unlearning [57.081646768835704]
  User specifications or legal frameworks often require information to be removed from pretrained models, including large language models (LLMs)<n>This requires deleting or "forgetting" a set of data points from an already-trained model, which typically degrades its performance on other data points.<n>We propose UPCORE, a method-agnostic data selection framework for mitigating collateral damage during unlearning.
  arXiv  Detail & Related papers  (2025-02-20T22:51:10Z)
• RealEra: Semantic-level Concept Erasure via Neighbor-Concept Mining [25.769144703607214]
  Concept erasure has been proposed to remove the model's knowledge about protected and inappropriate concepts. We propose RealEra to address this "concept residue" issue. We show that RealEra outperforms previous concept erasing methods in terms of superior erasing efficacy, specificity, and generality.
  arXiv  Detail & Related papers  (2024-10-11T17:55:30Z)
• TaCo: Targeted Concept Erasure Prevents Non-Linear Classifiers From Detecting Protected Attributes [4.2560452339165895]
  Targeted Concept Erasure (TaCo) is a novel approach that removes sensitive information from final latent representations. Our experiments show that TaCo outperforms state-of-the-art methods.
  arXiv  Detail & Related papers  (2023-12-11T16:22:37Z)
• Implicit Concept Removal of Diffusion Models [92.55152501707995]
  Text-to-image (T2I) diffusion models often inadvertently generate unwanted concepts such as watermarks and unsafe images. We present the Geom-Erasing, a novel concept removal method based on the geometric-driven control.
  arXiv  Detail & Related papers  (2023-10-09T17:13:10Z)
• Deep Concept Removal [29.65899467379793]
  We address the problem of concept removal in deep neural networks. We propose a novel method based on adversarial linear classifiers trained on a concept dataset. We also introduce an implicit gradient-based technique to tackle the challenges associated with adversarial training.
  arXiv  Detail & Related papers  (2023-10-09T14:31:03Z)
• LEACE: Perfect linear concept erasure in closed form [97.78661458934953]
  Concept erasure aims to remove specified features from an embedding.<n>We introduce LEAst-squares Concept Erasure (LEACE), a closed-form method which provably prevents all linear classifiers from detecting a concept while changing the embedding as little as possible.<n>We apply LEACE to large language models with a novel procedure called "concept scrubbing," which erases target concept information from every layer in the network.
  arXiv  Detail & Related papers  (2023-06-06T16:07:24Z)
• Trust your neighbours: Penalty-based constraints for model calibration [19.437451462590108]
  We present a constrained optimization perspective of SVLS and demonstrate that it enforces an implicit constraint on soft class proportions of surrounding pixels. We propose a principled and simple solution based on equality constraints on the logit values, which enables to control explicitly both the enforced constraint and the weight of the penalty.
  arXiv  Detail & Related papers  (2023-03-11T01:10:26Z)
• Active Refinement for Multi-Label Learning: A Pseudo-Label Approach [84.52793080276048]
  Multi-label learning (MLL) aims to associate a given instance with its relevant labels from a set of concepts. Previous works of MLL mainly focused on the setting where the concept set is assumed to be fixed. Many real-world applications require introducing new concepts into the set to meet new demands.
  arXiv  Detail & Related papers  (2021-09-29T19:17:05Z)
• Weakly-Supervised Cross-Domain Adaptation for Endoscopic Lesions Segmentation [79.58311369297635]
  We propose a new weakly-supervised lesions transfer framework, which can explore transferable domain-invariant knowledge across different datasets. A Wasserstein quantified transferability framework is developed to highlight widerange transferable contextual dependencies. A novel self-supervised pseudo label generator is designed to equally provide confident pseudo pixel labels for both hard-to-transfer and easy-to-transfer target samples.
  arXiv  Detail & Related papers  (2020-12-08T02:26:03Z)
• Supercharging Imbalanced Data Learning With Energy-based Contrastive Representation Transfer [72.5190560787569]
  In computer vision, learning from long tailed datasets is a recurring theme, especially for natural image datasets. Our proposal posits a meta-distributional scenario, where the data generating mechanism is invariant across the label-conditional feature distributions. This allows us to leverage a causal data inflation procedure to enlarge the representation of minority classes.
  arXiv  Detail & Related papers  (2020-11-25T00:13:11Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.